TABLE X.
Optimal Hyperparameters for MGH NeuroBank Corpus Level 4
| GCNN | regularization | 2.65e–2 |
| decay_steps | 410 | |
| learning_rate | 1.01e–2 | |
| pool | mpool1 | |
| momentum | 8.14e–1 | |
| num_epochs | 350 | |
| batch_size | 25 | |
| M | [138, 60] | |
| ps | [[2]] | |
| decay_rate | 9.98e–1 | |
| Ks | [[26]] | |
| Fs | [[31]] | |
| dropout | 6.22e–1 | |
| FF-ANN | hidden_layer_sizes | [976] |
| alpha | 1.16 | |
| power_t | 3.21e–1 | |
| activation | relu | |
| learning_rate_init | 4.05e–1 | |
| early_stopping | False | |
| momentum | 9.07e–1 | |
| tol | 1.00e–5 | |
| nesterovs_momentum | True | |
| learning_rate | invscaling | |
| KNNs | n_neighbors | 6 |
| metric | canberra | |
| weights | distance | |
| Linear Classifier | learning_rate | invscaling |
| tol | 1.00e–5 | |
| n_jobs | −1 | |
| power_t | 1.84e–1 | |
| penalty | l1 | |
| eta0 | 3.17e–4 | |
| loss | log | |
| l1_ratio | 4.06e–1 | |
| alpha | 1.23e–3 | |
| Random Forest | max_depth | 25 |
| max_leaf_nodes | 500 | |
| min_weight_fraction_leaf | 4.33e–4 | |
| min_samples_split | 2 | |
| min_samples_leaf | 1 | |
| n_estimators | 411 | |
| criterion | gini | |
| min_impurity_decrease | 3.64e–5 | |
| Decision Tree | min_samples_split | 2 |
| max_leaf_nodes | None | |
| criterion | gini | |
| min_impurity_decrease | 7.73e–5 | |
| min_weight_fraction_leaf | 2.53e–3 | |
| min_samples_leaf | 1 | |
| max_features | 250 | |
| max_depth | 100 | |
| splitter | best |